KR20100071554A - Actuator fault diagnosis of uavs using adaptive unknown input observers - Google Patents

Abstract

PURPOSE: A failure diagnostic method of a control surface driver of a drone aircraft using an observation plane is provided to sense the location of the control surface where a malfunction has occurred. CONSTITUTION: A failure diagnostic method of a control surface driver of a drone aircraft using an observation plane comprises: a step of estimating the direction and speed of wind which affects the flying of an aircraft(S2); a step of detecting trouble on a control surface by comparing an estimation output vector and a pre-input model output vector(S4); a step of classifying the control surface with a malfunction by comparing the estimation output vector and the model output vector(S5); a step of estimating the fixed angle of the control surface when the malfunction occurs(S6); and a step of re-forming an algorithm in order to detect new malfunctions by reflecting the malfunction detected from a fault detection step(S7).

Description

Actuator fault diagnosis of UAVs using adaptive unknown input observers

The present invention relates to a method for diagnosing a control plane driver failure of an unmanned aerial vehicle using an adaptive unknown input observer, and more particularly, to recognize a control plane in which a failure occurs when a control plane driver failure occurs, as well as estimating a fixed angle to the survival and reliability of the unmanned aerial vehicle. An object of the present invention is to provide a method for diagnosing a failure of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer.

During operation of the aircraft, a failure occurs in the control plane driver of the aircraft, so that the control plane is fixed at a specific position and may not move. In general, in the case of the control plane failure, a study to enable the automatic control by hardware or software method is in progress.

The software method is preferred in that weight can be reduced, space can be used efficiently, and additional cost can be reduced compared to hardware methods.

If a failure occurs in the control plane driver of the aircraft, the failure detection to find out whether the failure occurred, the failure separation to find out which control failure occurred, and to re-form the algorithm according to the control plane where the failure occurred The process is necessary.

   One of the methods widely used for fault detection and fault isolation of such a control plane driver is by using an observer.

The observer means a technique implemented by software for estimating or observing the state of an aircraft using a mathematical dynamics model of a given aircraft.

1 is a block diagram illustrating an algorithm of a control plane driver failure diagnosis method of an unmanned aerial vehicle according to the prior art.

The control plane driver failure diagnosis method according to the prior art includes a failure determination step of determining whether a control plane is broken by comparing a model output vector modeling a failure occurrence to which the control plane driver is fixed and an estimated output vector by an observer. .

That is, when a failure occurs in which the control surface is fixed, the residual, which is the difference between the model output vector and the estimated output vector by the observer, becomes 0, thereby recognizing the failure.

However, in the case of the unmanned aerial vehicle's control plane driver failure diagnosis method, it is difficult to successfully perform the reforming control because it cannot determine only the failure of the control plane or cannot estimate the fixed angle when a failure occurs when two or more control planes are fixed. Is most of them.

The present invention was devised to solve the above problems, it is possible to perform a robust diagnosis even if disturbances such as gusts during operation of the aircraft, it is possible to grasp the position of the two or more control surface failure occurs An object of the present invention is to provide a method for diagnosing a failure of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer.

The method for diagnosing a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention includes estimating a direction and a speed of a gust when a gust that interferes with normal operation of the aircraft occurs during operation of the aircraft; Model a failure occurrence to which the control plane driver is fixed, and compare a previously inputted model output vector and an estimated output vector by an unknown input observer according to the estimated direction and speed of the wind gust to determine whether one of the control planes has failed A fault detection step of determining; A fault separation step of comparing the model output vector and the estimated output vector to distinguish a control surface in which a failure occurs among the plurality of control surfaces; A failure location estimating step of estimating an angle at which the control surface on which the failure has occurred is fixed through an adaptive technique; And a reforming step of reforming the algorithm to detect a new failure by reflecting the failure detected in the failure detection step.

The fault detection step may be determined as a failure when the residual, which is the difference between the model output vector and the estimated output vector on at least one control surface of the plurality of control surfaces is 0, and the residual is not 0 on the remaining control surfaces.

In the fault location determining step, the control surface having the residual value of 0 may be determined as a fault location.

The unknown input observer may be formed by Equation 7, and the adaptive technique may be made by Equation 8.

[Equation 7]

[Equation 8]

.

.

In the control plane driver failure diagnosis method of the unmanned aerial vehicle using the adaptive unknown input observer according to the present invention, the survivability and reliability of the unmanned aerial vehicle can be improved by estimating a fixed angle within 1 second through the adaptive unknown input observer.

The failure detection method of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention has an advantage of determining whether the control plane is broken and the position of the control plane where the failure occurs even when a disturbance such as a gust is generated.

In addition, the control plane driver failure diagnosis method of the unmanned aerial vehicle using the adaptive unknown input observer according to the present invention, there is an advantage that it is possible to quickly determine the location of the control plane failure even if a failure occurs in two or more control surfaces.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, terms or words used in this specification and claims should not be construed in a common or dictionary sense, and the inventors will be required to properly define the concepts of terms in order to best describe their invention. Based on the principle that it can, it should be interpreted as meaning and concept corresponding to the technical idea of the present invention.

Therefore, the embodiments described in the specification and the drawings shown in the drawings are only the most preferred embodiment of the present invention and do not represent all of the technical idea of the present invention, various equivalents may be substituted for them at the time of the present application. It should be understood that there may be water and variations.

2 is a block diagram showing an algorithm of a control plane driver failure diagnosis method for an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention, Figure 3 is a control plane driver failure diagnosis method of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention The block diagram shows the unknown input observer and the adaptive technique.

In the embodiment of the present invention, when a failure occurs in which the i-th control plane driver is fixed, the linear model of the aircraft is represented by Equation 1.

[Equation 1]

는 B의 i번째 열을 0으로 한 B 행렬,

는 B의 i번째 열벡터,

는 i번째 입력 즉, 고장에 해당되는 고정된 각도이다.Where x (t) is a state variable, u (t) is a control input, d (t) is a disturbance such as a gust, y (t) is an output vector (measured value), and A, B, and C are system matrices. Is the disturbance variance matrix, i = 1, ..., p is the i-th driver failure,

Is a B matrix with the i th column of B as 0,

Is the i-th column vector of B,

Is the fixed angle corresponding to the i th input, that is, the fault.

In the present embodiment, y (t) is a model output vector modeling the occurrence of the failure that the control plane driver is fixed. Further, in E, the direction and speed of the gust are input.

를 추정한 추정치, 즉 조종면의 고정각도 추정치(

)를 구하는 것이 고장진단의 최종목적이 된다.On the other hand, the fixed position is actually unknown

Is an estimate of, ie a fixed angle estimate of the control plane (

) Is the final purpose of troubleshooting.

를 구하기 전 조종면 구동기의 고장여부를 판단하는 고장검출단계 와 고장 난 조종면 구동기의 위치를 알아내는 고장분리단계가 수행되어야 한다.remind

The fault detection step for determining the failure of the control plane driver and the fault isolation step for locating the failing control plane driver should be performed before obtaining.

의 차인 잔차

만 0인 경우에 조종면 구동기의 고장으로 판단한다.That is, in the fault detection step, the model output vector y (t) modeling the occurrence of the failure to which the control plane driver is fixed and the estimated vector by the unknown input observer

Residuals

If 0, it is regarded as a failure of the control plane driver.

In addition, the fault separation step is to determine the location of the faulty driver by using only the unknown input observer, the residual generated by the observer for the faulty driver becomes zero, and the residual for the remaining driver does not become zero.

The state equation for designing the observer is shown in [Equation 2].

[Equation 2]

[Equation 2] can be seen that when the i-th driver is broken [Equation 1].

A commonly used unknown input observer is shown in Equation 3 below.

&Quot; (3) "

: 관측기의 상태변수(observer state vector),

: Observer state vector,

: 상태변수의 추정치(estimated state vector).

: Estimated state vector.

+

)이다.Where F, G, K, and H are matrices designed such that the estimation error is zero, and K = (

+

)to be.

)는 [수학식 4]로 전개됨을 알 수 있다.Estimated Error (Equation 2) and [Equation 3]

) Can be seen as developed in [Equation 4].

[Equation 4]

는 [수학식 5]와 같다.In other words, the estimation error

Is the same as [Equation 5].

[Equation 5]

Here, it can be seen from Equation 5 that the estimation error of the observer is independent of the state variable, the output y (t) and the disturbance vector.

는

이므로 우변의 두 번째 항은 사라지고,

는 실제로 고정된 입력

가 되며 우변 끝항

는 추정치

가 된다. 상기 F의 특이치를 모두 음이 되도록 설계하고,

를 추정하면 i번째 조종면 구동기의 잔차

는 0이 된다.For the i-th driver failure in [Equation 5]

Is

, So the second term on the right side disappears,

Is actually a fixed input

To the right end term

Is an estimate

Becomes Design all the singular values of F to be negative,

Is estimated, the residual of the i th steering wheel driver

Becomes zero.

The fault isolation is performed here because the residual for the other control plane driver is not zero.

는 [수학식 6]에 의해 0이 아니다.That is, the residual of the j-th control surface driver substituted j in [Equation 5]

Is not 0 by [Equation 6].

&Quot; (6) "

와

는 같아질 수 있지만, 고장에 해당하는 고정된 값

와 실제 명령

는 같아질 수 없으므로

는 0이 될 수 없다. That is, for a fault that the i th control plane driver

Wow

Can be equal, but a fixed value corresponding to a failure

And the actual command

Cannot be equal to

Cannot be zero.

를 추정하면,

만이 0이 되어 고장검출 및 고장분리가 가능하다.therefore

Is estimated,

Only 0 is possible for fault detection and fault isolation.

를 이용한

를 구하는 방법을 설명한다.Next, residuals through adaptive techniques

Using

How to obtain.

That is, the adaptive unknown input observer according to the present invention is completed through the unknown input observer and the adaptive technique.

는 모르는 값이므로

을 [수학식 3]에 대입하면 다음 [수학식 7]과 같다.remind

Is an unknown value

Substituting into [Equation 3] is the following [Equation 7].

[Equation 7]

가 0이 되도록 설계되는 행렬이다.F, G, K, H is the estimated error

Is a matrix designed to be zero.

는 다음 [수학식 8]의 적응기법을 통해 구할 수 있다.Fixed angle estimate of the control surface

Can be obtained through the adaptation method of Equation 8.

[Equation 8]

는 수렴속도에 영향을 미치는 양수이고, C는 출력행렬, G는 상기 추정오차가 0이 되도록 설계되는 행렬,

는 잔차이며,

는 시스템행렬 B의 i번째 열벡터이고, P는 리아프노프 방정식, 즉 [수학식 9]를 만족시키는 대칭 양한정 행렬이다. remind

Is a positive number affecting the convergence speed, C is an output matrix, G is a matrix designed such that the estimation error is zero,

Is the residual,

Is the i-th column vector of the system matrix B, and P is a symmetric positive limiting matrix that satisfies the Lyapunov equation, [9].

[Equation 9]

는 다음 [수학식 10]과 같이 정의된다.Q is also an arbitrary symmetric definite matrix, the residual

Is defined as in Equation 10 below.

[Equation 10]

의 도함수는 다음 [수학식 11]과 같다.From Equation 5 above

The derivative of is given by Equation 11 below.

[Equation 11]

는

에 대한 추정오차이다.

Is

Estimated error for.

와

는 지수적으로 수렴됨이 보장되어야 하며, 이를 위해 다음 [수학식 12]와 같은 리아노프함수를 도입하여

와

의 수렴 안정성을 보인다.Where above

Wow

Should be guaranteed to be exponentially converged. For this purpose, we introduce the Lyanov function

Wow

Shows convergence stability.

[Equation 12]

을 이용하면

는 [수학식 13]과 같다.Equation 10, Equation 11 and

If you use

Is the same as [Equation 13].

[Equation 13]

Substituting Equation 8 into Equation 13, the following Equation 14 is obtained.

[Equation 14]

역시 대칭 양한정 행렬이 되고, F의 특이치가 모두 음이므로 다음의 리아프노프 방정식 즉, [수학식 15]를 만족하는 대칭 향한정 행렬

가 존재한다.Since rank (C) = n,

Again, it becomes a symmetric positive limiting matrix, and since all of the singular values of F are negative, the symmetric heading matrix satisfying the following Liafnov equation,

Is present.

[Equation 15]

Thus, the convergence stability of the adaptive unknown input observer according to the present invention has been demonstrated.

와

의 지수적 수렴성이 증명된다.By the following equation (16) for the next linear time-varying system

Wow

Exponential convergence is demonstrated.

[Equation 16]

Where F is Hurwitz and M (t) is bounded & Globally Lipschitz.

Error equations from Equations 8 and 11 are expressed as Equation 17 below.

[Equation 17]

와

는 0에 지수적으로 수렴됨이 증명되었다.Equation 17 corresponds to the form of Equation 16.

Wow

Proved to converge exponentially to zero.

Referring to the effect of the control plane failure diagnosis method of the aircraft according to the present invention configured as described above are as follows.

4 is a flowchart illustrating a method for diagnosing a failure of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention.

The method for diagnosing a failure of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention includes estimating the direction and speed of the gust when a gust that interferes with normal operation of the aircraft occurs during operation of the aircraft (S2) and Modeling a failure occurrence of the control plane driver is fixed, and compares the input model output vector and the estimated output vector by the unknown input observer according to the estimated direction and speed of the wind gusts to determine whether one of the control surfaces is a failure A fault detection step (S4) for determining a fault, a fault separation step (S5) for distinguishing a control surface in which a failure occurs among the plurality of control surfaces by comparing the model output vector and the estimated output vector, and the failure through an adaptive technique. The fault position estimating step (S6) of estimating the angle at which the generated control surface is fixed and the fault detected in the fault detecting step are reflected. And a reforming step (S7) of reforming the algorithm so that a new failure can be detected.

The fault detection step (S4) determines that a failure is obtained when the residual, which is the difference between the model output vector and the estimated output vector on at least one of the control surfaces, is zero, and the residual is not zero on the remaining control surfaces.

In the fault separation step, the control surface having the residual value of 0 is determined as a fault position.

와

가 지수적으로 수렴함을 통해 고장에 의해 고정된 조종면과 조종면의 고정각도를 추정할 수 있다.That is, through the above [Equation 12] to [Equation 17]

Wow

The exponential convergence makes it possible to estimate the control plane fixed by the fault and the fixed angle of the control plane.

When the fault detection step (S4), the fault separation step (S5), the fault location estimation step (S6) and the remodeling step (S7) is completed as described above, the observer structure changing step of changing the structure of the observer according to the failed control plane ( S8) may be further included.

As described above, although the present invention has been described by way of limited embodiments and drawings, the present invention is not limited thereto and is intended by those skilled in the art to which the present invention pertains. Of course, various modifications and variations are possible within the scope of equivalents of the claims to be described.

1 is a block diagram showing an algorithm of a control plane driver failure diagnosis method of an unmanned aerial vehicle according to the prior art,

2 is a block diagram showing an algorithm of a control plane driver failure diagnosis method of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention;

3 is a block diagram illustrating an adaptive unknown input observer of a method for diagnosing a control plane driver failure of an unmanned aerial vehicle using the adaptive unknown input observer according to the present invention.

4 is a flowchart illustrating a method for diagnosing a failure of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer according to the present invention.

Claims (4)

Estimating the direction and speed of the gust when a gust that interferes with normal operation of the aircraft occurs during operation of the aircraft; Model a failure occurrence to which the control plane driver is fixed, and compare a previously inputted model output vector and an estimated output vector by an unknown input observer according to the estimated direction and speed of the wind gust to determine whether one of the control planes has failed A fault detection step of determining; A fault separation step of comparing the model output vector and the estimated output vector to distinguish a control surface in which a failure occurs among the plurality of control surfaces; A failure location estimating step of estimating an angle at which the control surface on which the failure has occurred is fixed through an adaptive technique; And And a reforming step of reforming an algorithm to detect a new failure by reflecting the failure detected in the failure detection step. The method according to claim 1, The fault detection step, Using an adaptive unknown input observer, characterized in that the failure is determined when the residual which is the difference between the model output vector and the estimated output vector is zero on at least one of the control surfaces and the residual is not 0 on the remaining control surfaces. How to troubleshoot control plane driver of unmanned aerial vehicle. The method according to claim 1, The fault separation step, And a control plane driver failure diagnosis method of an unmanned aerial vehicle using an adaptive unknown input observer, characterized in that the control plane having the residual value of zero is determined as a failure location. The method according to claim 1, The unknown input observer is made of [Equation 7], The adaptive technique is a failure diagnosis method of a control plane driver of an unmanned aerial vehicle using an adaptive unknown input observer, characterized in that [Equation 8]. [Equation 7]

[Equation 8]

.

Images